When multiple threads call RdKafka::Headers at high concurrency, will the cpu utilization be low because of malloc?

[christian600]

I created a producer for each thread. Through RdKafka: :Producer->produce() when sending a message will bring RdKafka: : Headers. When my data volume is large, I want to add additional threads and producers to reduce latency. However, the effect is not ideal, and the reduction of delay by one more producer is limited.
At this point I see that RdKafka::Headers involves a malloc operation. Does this cause frequent context switching and memory fragmentation in multithreaded scenarios?
ldd (GNU libc) 2.17